Exploring the impact of GenAI-assisted EFL courses on foreign language anxiety and willingness to communicate: a disciplinary comparison

Abstract

Abstract The integration of generative artificial intelligence (GenAI) in language education has shown positive effects on language skills, anxiety reduction, and motivation. Yet, limited research has examined its differential impacts on foreign language anxiety (FLA) and willingness to communicate (WTC) among learners from varied academic disciplines. To fill the gap, this mixed-methods study compared universities students in Accounting (ACC) and Electrical and Electronic Engineering (EEE). With data from draw-a-picture tasks, questionnaires, and semi-structured interviews, results supported that GenAI-assisted intervention significantly reduced FLA and enhanced initial WTC, yet did not exert a significant influence on deeper levels of WTC. No statistically significant differences were observed between the two groups concerning changes in FLA and WTC. Qualitatively, we adopted an innovative analytical framework, T-CADS-GPT, which integrated Non-Negative Matrix Factorization with Fairclough’s three-dimensional discourse analysis model, to conduct fine-grained analysis of the interview transcripts. Interview data revealed that ACC students tended to search GenAI’s personalized learning support, whereas EEE students viewed GenAI more as a functional tool. The findings recommend that EFL instructors offer tailored GenAI guidance and technical assistance to students from diverse academic backgrounds, and integrate English for Specific Purposes content into curriculum design to fully harness the pedagogical affordances of GenAI.